Name
Abstract | ||
|---|---|---|
Discriminative training has received a lot of attention from both the machine learning and speech recognition communities. The idea behind the discriminative approach is to construct a model that distinguishes correct samples from incorrect samples, while the conventional generative approach estimates the distributions of correct samples. We propose a novel discriminative training method and apply it to a language model for reranking speech recognition hypotheses. Our proposed method has round-robin duel discrimination (R2D2) criteria in which all the pairs of sentence hypotheses including pairs of incorrect sentences are distinguished from each other, taking their error rate into account. Since the objective function is convex, the global optimum can be found through a normal parameter estimation method such as the quasi-Newton method. Furthermore, the proposed method is an expansion of the global conditional log-linear model whose objective function corresponds to the conditional random fields. Our experimental results show that R2D2 outperforms conventional methods in many situations, including different languages, different feature constructions and different difficulties. |
Year | DOI | Venue |
|---|---|---|
2012 | 10.1109/TASL.2011.2174225 | IEEE Transactions on Audio, Speech & Language Processing |
Keywords | Field | DocType |
error rate,conditional random fields,language model,quasi newton method,machine learning,speech,conditional random field,vectors,error correction,natural language processing,speech recognition,objective function,speech processing,parameter estimation,data models,learning artificial intelligence,log linear model | Conditional random field,Speech processing,Data modeling,Pattern recognition,Computer science,Word error rate,Speech recognition,Artificial intelligence,Estimation theory,Sentence,Discriminative model,Language model | Journal |
Volume | Issue | ISSN |
20 | 4 | 1558-7916 |
Citations | PageRank | References |
6 | 0.43 | 21 |
Authors | ||
4 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Takanobu Oba | 1 | 53 | 12.09 |
| Takaaki Hori | 2 | 408 | 45.58 |
| Atsushi Nakamura | 3 | 96 | 7.76 |
| Akinori Ito | 4 | 272 | 62.32 |